[MooseFS-Users] MFS 3.0 Internal Rebalancing vs Undergoal Chunks

[Michael T. <mic...@ho...>]

Hi.


I recently upgraded my setup from 2.0 to 3.0.  My master, all metaloggers, and all chunkservers are already upgraded.  Most of my clients are still on 2.0 and will be upgraded this weekend.


A couple of days ago, one of the chunkservers had a disk start to fail.  So I marked it for removal and restarted the chunkserver service, and MFS started removing th chunks.  Not long after, the disk totally failed, but the removal of the chunks have not finished.  No big deal though.


So I replaced the disk with a larger sized one: old=2TB, new=4TB.  Restarted the chunkserver and the new disk shows up in the CGI GUI.  24 hours later, however, I noticed that the number of undergoal chunks did not seem to shrink much.  Looking at the logs, it seems that Internal Rebalancing (e.g. moving chunks from other disks into the new one) seems to be taking priority over reaching goal for undergoal chunks.


How can I get the master to do the reverse, i.e. prioritize reaching the goal for undergoal chunks over internal rebalancing?



--- mike t.